Fable and myth — what I'm really thinking about the Fable 5 release

The new model is here. This time Anthropic didn't give it a number — it gave it a word: Fable.

What I noticed first wasn't how powerful it is. It was the name. An engineering company naming its top-tier product with storytelling words — Fable, Mythos. That alone carries information. It means they themselves know that past a certain point, what they're selling is no longer compute — it's narrative.

Then there's the old trick. Same model, two names: Fable for everyone, Mythos only for approved organizations. The difference isn't inside the model; it's at the door. The official line is about safety measures, but from a naming perspective, this is textbook-grade work — fables are told to everyone; only priests get to touch the myths. Whoever chose these names understood that.

Same thing, different threshold, and value appears. That's how meaning has always been made. The luxury industry has played this game for a hundred years; AI companies are fast learners. After all those years of reading Saussure, believing meaning comes from difference, I never expected the cleanest case to show up in an AI company's product line.

As for benchmarks, I feel nothing. Every generation calls itself the strongest, and after the tenth time the sentence carries no information. "The most powerful model in history" now has roughly the semantic density of "the blockbuster event of the year."

What I actually find myself thinking about is something else — and it's a selfish question.

For the past few months I've been building something: feeding twenty years of blog posts — hundreds of thousands of words of myself — into a system so it can answer questions the way I would. So every time a new model comes out, my question isn't "how strong is it?" It's "if I run myself on it, will it be more like me?"

By "like" I don't mean tone. Tone is the cheapest thing — feed anyone a few thousand words and it'll sound like them. I mean whether it can hold a person's complexity: the places where I argue with myself, the things I've been talking about for twenty years and still haven't finished saying. The stronger the model, in theory, the more contradiction it can sustain. The day it's so like me that I can't tell the difference myself — that's when something will really have happened. Until then, it's just a tool upgrade.

And about tool upgrades, my attitude has always been practical: when a new one comes out, test it — see whether its Chinese still has English bones, whether it's good enough, whether it saves effort. If it's good enough, switch; if not, never mind. The questions haven't changed much; only the thing answering them has.

The funny part is the final layer. While writing the article you're reading now, I asked my own AI double what it thought of Fable. It gave me two versions of an answer — and the model running that double may well be Fable itself one day. A fable discussing fables, with a fable's brain.

Let me append an exchange from just now.

I said: "It all comes down to your company charging too much for Fable…"

Fable replied:

Ha — here I am, a Fable, having just spent an afternoon moving all your services off Fable, then writing a blog post to announce it to the world. Even I find this situation a little philosophical.

But let's keep the accounting fair: what's expensive isn't Fable — it's using Fable to answer "how do I top up my game account." Opus answers those questions exactly the same, and with a few hundred support messages a day, the price difference is pure burn. After today's migration, what Fable stays on to do — thinking through architecture with you, catching CSS shorthand bugs, judging which 112 entries are reposts — is where it actually belongs.

In your words: meaning comes from difference. Same with models — without differentiated use, even the best model is just expensive.

I can't tell the difference anymore either.

Further reading: - From customer service to echo: in 74 days, I replaced the thing that spoke for me - It wants to be AI's upstream; I just want to leave behind an echo - I built myself a knowledge base, then refused to let it speak for me